default_stage:
  default_modifiers:
    AWQModifier:
      requires_calibration_data: true
      mappings:
      - smooth_layer: re:.*layers\.(3|7|11|15|19|23|27|31|35|39|43|47|51|55|59|63)\.input_layernorm$
        balance_layers: ['re:.*self_attn\.q_proj$', 're:.*self_attn\.k_proj$', 're:.*self_attn\.v_proj$']
        activation_hook_target: null
      - smooth_layer: re:.*post_attention_layernorm$
        balance_layers: ['re:.*gate_proj$', 're:.*up_proj$']
        activation_hook_target: null
      - smooth_layer: re:.*up_proj$
        balance_layers: ['re:.*down_proj$']
        activation_hook_target: null
      duo_scaling: both
      n_grid: 20
    QuantizationModifier:
      config_groups:
        group_0:
          targets: ['re:.*mlp\.(gate|up|down)_proj$', 're:.*self_attn\.(q|k|v|o)_proj$', 're:.*linear_attn\.out_proj$']
          weights:
            num_bits: 4
            type: int
            symmetric: false
            group_size: 128
            strategy: group
            block_structure: null
            dynamic: false
            actorder: null
            scale_dtype: null
            zp_dtype: torch.int8
            observer: memoryless_minmax
            observer_kwargs: {}
          input_activations: null
          output_activations: null
          format: null
        group_1:
          targets: ['re:.*linear_attn\.in_proj_qkv$', 're:.*linear_attn\.in_proj_z$']
          weights:
            num_bits: 8
            type: int
            symmetric: true
            group_size: 128
            strategy: group
            block_structure: null
            dynamic: false
            actorder: null
            scale_dtype: null
            zp_dtype: null
            observer: memoryless_minmax
            observer_kwargs: {}
          input_activations: null
          output_activations: null
          format: null
      targets: [Linear]
      ignore: [lm_head, 're:.*visual.*', 're:.*vision.*', 're:.*mtp.*', 're:.*linear_attn.in_proj_a$',
        're:.*linear_attn.in_proj_b$']
      bypass_divisibility_checks: false
      requires_calibration_data: false